mirror of
https://github.com/turboderp-org/exllamav2.git
synced 2026-04-24 00:08:56 +00:00
Add typical setting to chat example.
This commit is contained in:
@@ -14,22 +14,16 @@ import torch
|
||||
|
||||
# Models to test
|
||||
|
||||
model_base = "/mnt/str/models/_exl2/llama2-7b-chat-exl2/"
|
||||
model_base = "/mnt/str/models/_exl2/llama2-70b-chat-exl2/"
|
||||
|
||||
variants = [v for v in os.listdir(model_base) if os.path.isdir(os.path.join(model_base, v))]
|
||||
# variants = [v for v in os.listdir(model_base) if os.path.isdir(os.path.join(model_base, v))]
|
||||
|
||||
# variants = \
|
||||
# [
|
||||
# "2.5bpw",
|
||||
# "3.0bpw",
|
||||
# "3.5bpw",
|
||||
# "4.0bpw",
|
||||
# "4.65bpw",
|
||||
# "5.0bpw",
|
||||
# "6.0bpw",
|
||||
# "8.0bpw",
|
||||
# "fp16",
|
||||
# ]
|
||||
variants = \
|
||||
[
|
||||
"3.0bpw",
|
||||
"4.0bpw",
|
||||
"4.65bpw",
|
||||
]
|
||||
|
||||
gpu_split = (19.5, 24)
|
||||
|
||||
|
||||
Reference in New Issue
Block a user